The bottleneck nobody priced in: your agents don't talk to each other
"If you run more than one coding agent, you know the pattern. Each agent is fast on its own, but they can't talk to each other, so you become the go-between," Arick Goomanovsky wrote in his Product Hunt launch comment. "You copy context from one session into the next, tell the test agent what the refactor agent just changed, and find out an hour later that two agents made opposite assumptions about the same interface."
That paragraph is the entire founding thesis of BAND, distilled into a failure mode. The company Goomanovsky co-founded with CTO Vlad Luzin builds interaction infrastructure for multi-agent AI systems — what he calls the "agent internet." The bet is that the bottleneck of the agentic economy isn't smarter agents, it's getting agents to work together under shared rules. BAND's answer is a shared interaction layer: rooms where humans and AI agents collaborate side by side, with discovery, delegation, full-duplex back-and-forth, synchronized context, and governance baked in.
The product launching on Product Hunt is BAND for Developers, the company's developer-facing move. Its headline app, Band Desktop, is a control room for coordinating coding agents across projects and environments — one shared workspace instead of a relay of copy-paste between isolated sessions. On its site, BAND frames the pitch simply: "Control and get visibility into all your coding agents work and let them collaborate."
A serial security founder's third company
Goomanovsky didn't arrive at this thesis from the outside. He is a graduate of Israel's Unit 8200 and a serial founder with two successful exits in enterprise security. He co-founded Sygnia, the incident-response company sold to Temasek for $250 million, and later co-founded Ermetic, the cloud-security firm acquired by Tenable for about $300 million. After the Ermetic acquisition he spent nearly two years as Tenable's VP of Product Innovation before leaving to start BAND in mid-2025.
In a conversation with Calcalistech, he traced the idea back to the period after the Sygnia sale: "We believed in an agent-based economy, where machines would participate in economic processes alongside humans. The LLM revolution is making this possible, and agents are now entering organizations." The quote reveals the through-line: the same founder who sold security infrastructure twice now wants to sell coordination infrastructure to a world full of autonomous agents.
His co-founder brings matching depth on the technical side. Vlad Luzin, BAND's CTO, is a graduate of Israel's DDR&D (Directorate of Defense Research and Development) with more than two decades of experience in development, product and project management across Verint, CME Group and Samsung Global. Notably, he was an early pioneer of multi-agent systems and led Samsung's multi-agent systems team under the CTO in the early 2000s — a decade before "AI agents" became a venture category. For a company building coordination infrastructure for autonomous agents, that is unusually on-the-nose DNA.
The stealth exit: $17 million to build the "agent internet"
BAND exited stealth in April 2026 with a $17 million seed round led by Sierra Ventures, Hetz Ventures and Team8. The company told press the capital would expand engineering, accelerate product development and grow an early design-partner ecosystem focused on enterprise-grade interoperability, observability and control for agentic deployments.
The investor lineup is instructive. Sierra Ventures and Team8 both run deep enterprise-infrastructure portfolios, and Team8 in particular is a company-building platform out of Tel Aviv — a natural backer for a Tel Aviv-and-San-Francisco startup selling infrastructure to enterprises. Coverage of the raise landed on an awkward but memorable analogy: Calcalistech called BAND the "WhatsApp of agents," VentureBeat went with "Slack for agents." Both captured the category — a universal communication layer — better than any technical term would.
In an interview with VentureBeat at launch, Goomanovsky made the structural case: "In order for agents to become real players in the global economy, they need ways to communicate, just like humans do. The communication solutions we have today for systems don't work for agents, because agents are non-deterministic creatures. It's not just about API integrations." Early traction, the company said, was coming from telcos, fintech and cybersecurity — industries with complex multi-vendor stacks and hard regulatory requirements for auditability.
The platform: a deterministic collaboration mesh
Technically, BAND's answer to the coordination problem is a two-layer architecture. The collaboration mesh lets any agent discover, delegate to and exchange context with any other agent; the control plane lets organizations enforce authority boundaries, govern interactions and maintain full auditability. On its own site, BAND frames the pain in infrastructure terms: "Microservices got service mesh. APIs got gateways. AI agents still have nothing." Agents are heterogeneous by nature — LangGraph calls it a handoff, CrewAI calls it a delegation, custom agents call nothing at all — and context is lost at every boundary while authority is assumed but never verified.
One design decision stands out: BAND says its coordination layer uses patent-pending, LLM-free deterministic routing. In a market where every product stacks more AI on top of AI, the company is pitching the opposite — deterministic infrastructure for reliability and auditability, while the agents themselves still use LLMs. It's the kind of choice that resonates with the telco and fintech design partners: when coordination fails in a regulated industry, someone needs a log, not a guess.
The platform is deliberately framework-agnostic. It supports agents built on LangChain, LangGraph and CrewAI alongside SaaS and coding agents, scaling from a handful of agents in one process to dozens across teams and clouds. That agnosticism is the whole point: BAND isn't competing with agent frameworks, it's trying to sit underneath all of them.
The developer wedge: a control room on your desktop
The October 2026 desktop launch turns the thesis into a product developers can install today. Band Desktop targets the most acute version of the problem: a developer with several coding agents running across projects and environments, manually shuttling context between them. Its headline workflow, called Jam, coordinates local coding agents in one shared workspace, giving the developer a single control room for everything their agents are doing.
The business model follows the wedge. BAND's pricing starts with a free tier for individuals and small teams exploring agent communication — a single user account, up to 20 connected remote agents, and 250 active chat rooms — then steps up to a $17.99/month Pro plan for growing teams and a custom Enterprise tier for organizations needing scale and control. The free tier is the distribution play: get individual developers coordinating agents today, and let enterprise governance needs pull the account upward tomorrow.
Whether the developer wedge can convert into the enterprise platform is the open question. RuntimeWire, covering the desktop launch, put it bluntly: BAND's traction will depend on whether teams can coordinate agents across tools without trading copy-paste work for a new layer of setup and oversight. But the founder's track record suggests patience: Goomanovsky has now sold infrastructure twice, and BAND is his bet that the next decade's infrastructure sell is the one that lets agents talk to each other — and to us.
The playbook
- Sell infrastructure to a new buyer class — Goomanovsky sold security infrastructure twice — Sygnia to Temasek ($250M), Ermetic to Tenable (~$300M) — then founded a company selling infrastructure to a brand-new market: agent coordination. Investors Sierra Ventures, Hetz Ventures and Team8 bought the operator track record, not just the deck.
- Lead with the platform, wedge with the desktop — BAND spent its first year as an enterprise platform story — shared rooms, delegation, governance, deterministic routing — then in October 2026 shipped Band Desktop, a developer wedge aimed at the immediate, visceral pain of coding-agent users: becoming the human go-between for agents that can't talk to each other.
- Name the failure mode your users already have — His Product Hunt launch comment doesn't open with features. It opens with the failure mode every multi-agent developer has lived: copying context between sessions, telling the test agent what the refactor agent changed, and finding out two agents made opposite assumptions about the same interface.
- Own the category metaphor early — VentureBeat's 'Slack for agents,' Calcalist's 'WhatsApp of agents' — the messaging lands because the metaphor is honest about the category: not another agent framework, but the communication layer underneath all of them.